betaFeedback

WorksMathematics

Total Variation Regularized Barycenters in Wasserstein Space

Abstract

We study total variation regularization of Wasserstein barycenters on bounded connected Lipschitz domains. We prove existence for arbitrary prescribed probability measures and uniqueness when at least one is absolutely continuous. Our main results are attained dual formulations in terms of relative perimeters and vector fields with zero normal trace. These yield necessary and sufficient optimality conditions and a variational characterization of superlevel sets. We also establish a maximum principle on convex domains, stability of minimizers, and an exact reduction to a single TV-JKO step in one dimension. The existence, uniqueness, and perimeter duality results extend to costs ∣x−y∣p|x-y|^p, p>1p>1, under the same assumptions.

Introduction

Wasserstein barycenters provide a notion of averaging probability measures that accounts for the geometry of their underlying space. Agueh and Carlier [1] developed their variational theory and established connections with multi-marginal optimal transport. Subsequent work has extended this framework to more general metric settings and studied its applications in imaging and statistics; see, for example, [17, 22, 6, 15, 26]. Regularization provides a way to impose additional structure on the barycenter. Bigot, Cazelles and Papadakis [5] studied penalized Wasserstein barycenters, while Carlier, Eichinger and Kroshnin [8] investigated entropy penalties on the barycenter density. Such penalties act on the unknown measure itself. A different approach regularizes the transport problems, for example by adding entropy to the transport plans; see [21]. Chizat [10] studies a doubly regularized formulation combining entropic transport costs with an entropy penalty on the barycenter, and establishes regularity and stability results. Our formulation retains the unregularized transport costs and penalizes the total variation of the barycenter density.

Let Ω⊂Rd\Omega\subset\mathbb{R}^{d} be a nonempty bounded connected Lipschitz domain, and let μ1,…,μn∈P(Ω‾)\mu_{1}, \ldots, \mu_{n} \in\mathcal{P}(\overline{\Omega}). We seek a barycenter whose density has bounded variation and penalize its spatial variation. To formulate this problem on P(Ω‾)\mathcal{P}(\overline{\Omega}), we extend the total variation of densities by +∞+\infty outside this class. For ν∈P(Ω‾)\nu\in\mathcal{P}(\overline{\Omega}), define

T(ν):={∣Dρ∣(Ω),if ν=ρ dx for some ρ∈BV(Ω),+∞,otherwise.\mathcal{T}(\nu) := \begin{cases} |D\rho|(\Omega), & \text{if } \nu=\rho\,dx \text{ for some } \rho\in BV(\Omega),\\ +\infty, & \text{otherwise.} \end{cases}

We consider

inf⁡ν∈P(Ω‾){∑i=1nW22(μi,ν)+T(ν)}.\inf_{\nu\in\mathcal{P}(\overline{\Omega})}\left\{\sum_{i=1}^{n} W_{2}^{2}(\mu_{i},\nu)+\mathcal{T}(\nu)\right\}.

2020 Mathematics Subject Classification. 49Q22, 49Q20, 26B30.

Key words and phrases. Wasserstein barycenter; total variation; convex duality; functions of bounded variation; relative perimeter. The probability measure with constant density 1/∣Ω∣1/|\Omega| on Ω\Omega has finite energy, so the infimum is finite. By the definition of T\mathcal{T}, the problem is therefore equivalent to minimization over probability densities in BV(Ω)BV(\Omega):

inf⁡ρ∈A{∑i=1nW22(μi,ρ dx)+∣Dρ∣(Ω)},A={ρ∈BV(Ω):ρ≥0 a.e., ∫Ωρ dx=1}.(1)\inf_{\rho\in\mathcal{A}}\left\{\sum_{i=1}^{n} W_2^2(\mu_i,\rho\,dx)+|D\rho|(\Omega)\right\}, \qquad\mathcal{A}=\left\{\rho\in BV(\Omega):\rho\geq0\ \text{a.e.},\ \int_{\Omega}\rho\,dx=1\right\}. \tag*{(1)}

The prescribed measures may be singular, whereas every minimizer has a density in BV(Ω)BV(\Omega). The total variation penalty controls spatial variation while allowing jump discontinuities. Its use in image restoration goes back to Rudin, Osher and Fatemi [23]. In the present setting, it controls the spatial variation of the barycenter without requiring its density to be smooth or strictly positive.

The closest related variational problem is the TV-JKO scheme studied by Carlier and Poon [9]. For a given probability measure μ\mu, one step of this scheme minimizes

12τW22(μ,ρ dx)+∣Dρ∣(Ω)\frac{1}{2\tau}W_2^2(\mu,\rho\,dx)+|D\rho|(\Omega)

over ρ∈A\rho\in\mathcal{A}. Thus (1.1) with n=1n=1 corresponds to τ=1/2\tau=1/2. We adapt the entropic approximation of Carlier and Poon [9] to derive vector-field optimality conditions for TV-regularized Wasserstein barycenters, allowing singular prescribed measures. We also extend their projection argument to obtain a maximum principle on convex domains. We further establish an attained dual formulation through relative-perimeter inequalities and relate it to the vector-field formulation. This duality characterizes almost every superlevel set of positive measure of a minimizing density as a minimizer of a perimeter ratio. In one dimension, an exact variance identity reduces the barycenter problem to a single TV-JKO step.

Main results

We first prove existence for arbitrary μ1,…,μn∈P(Ω‾)\mu_1,\ldots,\mu_n\in\mathcal{P}(\overline{\Omega}). If at least one of these measures is absolutely continuous, the classical strict-convexity argument for the transport cost gives uniqueness. We then establish an attained dual formulation involving transport potentials, a scalar multiplier, and relative-perimeter constraints.

The duality identity yields necessary and sufficient conditions for optimality. These conditions characterize optimal transport plans and imply that almost every superlevel set of positive measure of a minimizing density minimizes the relative-perimeter ratio associated with the sum of the optimal transport potentials. They also give a transport-map representation when a prescribed measure is absolutely continuous and a criterion for the uniform density to be optimal.

Following the entropy approximation in [9], we establish a second attained dual formulation in terms of bounded vector fields. The resulting optimality conditions retain the multiplier for the mass constraint and a nonnegative multiplier associated with the constraint ρ≥0\rho\geq0. This construction does not require the prescribed measures to be absolutely continuous.

We further prove a maximum principle on convex domains and stability under approximation of the prescribed measures. In one dimension, a variance identity reduces the problem exactly to a TV-JKO step with the ordinary Wasserstein barycenter as reference measure. Finally, existence, uniqueness under the same absolute-continuity condition, and the perimeter dual formulation extend to costs ∣x−y∣p|x-y|^p with p>1p>1.

Organization of the paper

Section 2 introduces the notation and recalls the analytical tools. Section 3 establishes existence, uniqueness, and the perimeter dual formulation. Section 4 derives the corresponding optimality conditions and transport representations. Section 5 develops the entropy approximation and proves vector-field duality. Section 6 contains the maximum principle, the one-dimensional reduction, and the stability result. Section 7 treats the extension to pp-Wasserstein costs.

Notation and preliminaries

We introduce the notation used throughout the paper and recall the compactness, optimal transport, and minimax results needed below.

Notation

Throughout, Ω⊂Rd\Omega\subset\mathbb{R}^{d}, d≥1d \ge1, is a nonempty bounded connected open set with Lipschitz boundary. Set K=Ω‾K = \overline{\Omega} and D=diam⁡(K)D = \operatorname{diam}(K). Convexity of Ω\Omega will be assumed only when explicitly stated. We denote by Ld\mathcal{L}^{d} the Lebesgue measure, and use ∣E∣|E| and 1E\mathbf{1}_{E} for the Lebesgue measure and characteristic function of a measurable set EE. The notation ∣⋅∣|\cdot| also denotes the Euclidean norm, with the meaning determined by the argument. Unless another measure is specified, almost-everywhere statements refer to Lebesgue measure.

For 1≤s<∞1 \le s < \infty, we use the norms

∥u∥Ls(Ω)=(∫Ω∣u∣s dx)1/s,∥u∥L∞(Ω)=ess sup⁡x∈Ω∣u(x)∣.\|u\|_{L^{s}(\Omega)} = \left(\int_{\Omega} |u|^{s}\,dx\right)^{1/s}, \qquad\|u\|_{L^{\infty}(\Omega)} = \operatorname*{ess\,sup}_{x \in\Omega} |u(x)|.

The same notation is used for vector-valued functions, with the Euclidean norm inside the integral. We write C(K)C(K) for the real-valued continuous functions on KK, equipped with the uniform norm

∥f∥C(K)=max⁡x∈K∣f(x)∣.\|f\|_{C(K)} = \max_{x \in K} |f(x)|.

The abbreviation ∥⋅∥∞\|\cdot\|_{\infty} denotes the uniform norm for functions in C(K)C(K) and the essential supremum norm for measurable functions on Ω\Omega. We use Cc1(Ω;Rd)C^{1}_{c}(\Omega;\mathbb{R}^{d}) for continuously differentiable vector fields with compact support in Ω\Omega. For a real-valued function uu, set

u+=max⁡{u,0},u−=max⁡{−u,0},u‾=1∣Ω∣∫Ωu dxif u∈L1(Ω).u_{+} = \max\{u,0\}, \qquad u_{-} = \max\{-u,0\}, \qquad\overline{u} = \frac{1}{|\Omega|}\int_{\Omega} u\,dx \quad\text{if } u \in L^{1}(\Omega).

The symbols →\to, ⇀\rightharpoonup, and ⇀∗\rightharpoonup^{*} denote strong, weak, and weak-star convergence in the indicated function spaces. For u∈L1(Ω)u \in L^{1}(\Omega), its distributional gradient is denoted by DuDu. Define

TV⁡(u)=sup⁡{∫Ωu div⁡z dx:z∈Cc1(Ω;Rd), ∥z∥∞≤1}.(2)\operatorname{TV}(u) = \sup\left\{\int_{\Omega} u\,\operatorname{div} z\,dx : z \in C^{1}_{c}(\Omega;\mathbb{R}^{d}),\ \|z\|_{\infty} \le1\right\}. \tag*{(2)}

where div⁡z=∑j=1d∂jzj\operatorname{div} z = \sum_{j=1}^{d} \partial_{j}z_{j}. Then BV⁡(Ω)={u∈L1(Ω):TV⁡(u)<∞}\operatorname{BV}(\Omega) = \{u \in L^{1}(\Omega) : \operatorname{TV}(u) < \infty\}. For u∈BV⁡(Ω)u \in\operatorname{BV}(\Omega), DuDu is a finite vector-valued Radon measure, and

TV⁡(u)=∣Du∣(Ω),∥u∥BV⁡(Ω)=∥u∥L1(Ω)+∣Du∣(Ω).\operatorname{TV}(u) = |Du|(\Omega), \qquad\|u\|_{\operatorname{BV}(\Omega)} = \|u\|_{L^{1}(\Omega)} + |Du|(\Omega).

Here ∣Du∣|Du| denotes the total variation measure of DuDu. For functions outside BV⁡(Ω)\operatorname{BV}(\Omega), the expression ∣Du∣(Ω)|Du|(\Omega) is understood as +∞+\infty. For a measurable set E⊂ΩE \subset\Omega, its relative perimeter is Per⁡(E;Ω)=∣D1E∣(Ω)\operatorname{Per}(E;\Omega) = |D\mathbf{1}_{E}|(\Omega). Thus only variation inside Ω\Omega is counted.

We write P(K)\mathcal{P}(K) for the Borel probability measures on KK. For p≥1p \ge1, let

Pp(K)={μ∈P(K):∫K∣x∣p dμ(x)<∞}.\mathcal{P}_{p}(K) = \left\{\mu\in\mathcal{P}(K) : \int_{K} |x|^{p}\,d\mu(x) < \infty\right\}.

Since KK is compact, Pp(K)=P(K)\mathcal{P}_p(K)=\mathcal{P}(K). For μ,ν∈P(K)\mu,\nu\in\mathcal{P}(K), the set Π(μ,ν)\Pi(\mu,\nu) consists of probability measures on K×KK\times K with first marginal μ\mu and second marginal ν\nu. For a Borel map TT, its pushforward is defined by (T#μ)(B)=μ(T−1(B))(T_{\#}\mu)(B)=\mu(T^{-1}(B)) for every Borel set BB in the target space. We denote the identity map by id⁡\operatorname{id} and the Dirac measure at xx by δx\delta_x. Narrow convergence μk⇀μ\mu_k\rightharpoonup\mu means

∫Kφ dμk→∫Kφ dμfor every φ∈C(K).\int_K \varphi\,d\mu_k\to\int_K \varphi\,d\mu\qquad\text{for every }\varphi\in C(K).

Optimal transport and analytical preliminaries

We first recall the compactness and coarea properties of BVBV functions, followed by the transport duality and minimax results used in the analysis.

On a bounded Lipschitz domain, every bounded sequence in BV(Ω)BV(\Omega) has a subsequence converging strongly in L1(Ω)L^1(\Omega) to a function in BV(Ω)BV(\Omega). Moreover,

uk→u in L1(Ω)⟹∣Du∣(Ω)≤lim inf⁡k→∞∣Duk∣(Ω).u_k\to u\text{ in }L^1(\Omega)\quad\Longrightarrow\quad|Du|(\Omega)\leq\liminf_{k\to\infty}|Du_k|(\Omega).

The embedding BV(Ω)↪Ls(Ω)BV(\Omega)\hookrightarrow L^s(\Omega) is compact for 1≤s<d/(d−1)1\leq s<d/(d-1) when d≥2d\geq2, and for every finite s≥1s\geq1 when d=1d=1. We also use the composition estimate ∣D(Φ∘u)∣(Ω)≤Lip⁡(Φ)∣Du∣(Ω)|D(\Phi\circ u)|(\Omega)\leq\operatorname{Lip}(\Phi)|Du|(\Omega) for u∈BV(Ω)u\in BV(\Omega) and Lipschitz functions Φ:R→R\Phi:\mathbb{R}\to\mathbb{R}. For nonnegative u∈BV(Ω)u\in BV(\Omega), the coarea and layer-cake formulas give

∣Du∣(Ω)=∫0∞Per⁡({u>t};Ω) dt.(3)|Du|(\Omega)=\int_0^\infty\operatorname{Per}(\{u>t\};\Omega)\,dt. \tag*{(3)}
∫Ωu dx=∫0∞∣{u>t}∣ dt.(4)\int_\Omega u\,dx=\int_0^\infty|\{u>t\}|\,dt. \tag*{(4)}

In particular, {u>t}\{u>t\} has finite relative perimeter for almost every t>0t>0. If V∈L∞(Ω)V\in L^\infty(\Omega), Fubini’s theorem also yields

∫ΩVu dx=∫0∞∫{u>t}V dx dt.\int_\Omega Vu\,dx=\int_0^\infty\int_{\{u>t\}}V\,dx\,dt.

We refer to [2] for these facts.

References

  1. [1]M. Agueh and G. Carlier, Barycenters in the Wasserstein space, SIAM J. Math. Anal. 43 (2011), 904–924.DOI
  2. [2]L. Ambrosio, N. Fusco and D. Pallara, Functions of Bounded Variation and Free Discontinuity Problems, Oxford University Press, Oxford, 2000.
  3. [4]G. Anzellotti, Pairings between measures and bounded functions and compensated compactness, Ann. Mat. Pura Appl. 135 (1983), 293–318.DOI
  4. [5]J. Bigot, E. Cazelles and N. Papadakis, Penalization of barycenters in the Wasserstein space, SIAM J. Math. Anal. 51 (2019), 2261–2285.DOI
  5. [6]N. Bonneel, J. Rabin, G. Peyré and H. Pfister, Sliced and Radon Wasserstein barycenters of measures, J. Math. Imaging Vision 51 (2015), 22–45.
  6. [7]C. Brizzi and L. Portinale, On the q-integrability of p-Wasserstein barycenters, preprint arXiv:2602.18293, 2026.arxiv.org/abs/2602.18293
  7. [8]G. Carlier, K. Eichinger and A. Kroshnin, Entropic-Wasserstein barycenters: PDE characterization, regularity, and CLT, SIAM J. Math. Anal. 53 (2021), 5880–5914.DOI
  8. [9]G. Carlier and C. Poon, On the total variation Wasserstein gradient flow and the TV-JKO scheme, ESAIM Control Optim. Calc. Var. 25 (2019), 42.arxiv.org/abs/1703.00243
  9. [10]L. Chizat, Doubly regularized entropic Wasserstein barycenter, Found. Comput. Math. 26 (2026), 2091–2128.DOI
  10. [11]G. De Philippis, A. R. Mészáros, F. Santambrogio and B. Velichkov, BV estimates in optimal transportation and applications, Arch. Ration. Mech. Anal. 219 (2016), 829–860.DOI
  11. [15]Y.-H. Kim and B. Pass, Wasserstein barycenters over Riemannian manifolds, Adv. Math. 307 (2017), 640–683.arxiv.org/abs/1412.7726
  12. [17]T. Le Gouic and J.-M. Loubes, Existence and consistency of Wasserstein barycenters, Probab. Theory Related Fields 168 (2017), 901–917.arxiv.org/abs/1506.04153
  13. [21]G. Peyré and M. Cuturi, Computational optimal transport, Found. Trends Mach. Learn. 11 (2019), 355–607.arxiv.org/abs/1803.00567
  14. [22]J. Rabin, G. Peyré, J. Delon and M. Bernot, Wasserstein barycenter and its application to texture mixing, in Scale Space and Variational Methods in Computer Vision, Lecture Notes in Computer Science, vol. 6667, Springer, Berlin, Heidelberg, 2012, pp. 435–446
  15. [23]L. I. Rudin, S. Osher and E. Fatemi, Nonlinear total variation based noise removal algorithms, Physica D 60 (1992), 259–268.DOI
  16. [24]F. Santambrogio, Optimal Transport for Applied Mathematicians: Calculus of Variations, PDEs, and Modeling, Progress in Nonlinear Differential Equations and Their Applications, vol. 87, Birkhäuser, Cham, 2015.DOI
  17. [25]M. Sion, On general minimax theorems, Pacific J. Math. 8 (1958), 171–176.DOI
  18. [26]S. Srivastava, C. Li and D. B. Dunson, Scalable Bayes via barycenter in Wasserstein space, J. Mach. Learn. Res. 19 (2018), 1–35.

Paper details

vlnc:W_8seszdjv
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 3

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents